Very Rough Software

This page contains links to some of the procedures I've written (mostly in Splus and/or R ) to make my life easier, or to implement methods that have not yet been integrated into the mainstream of data analysis packages. Everything here was written with a specific application in mind, and therefore will probably crash if you try to use it without changing the code yourself. In the future, I'll try to generalize these bits and pieces of code so that they are easier to use. For now, this page allows me to access some of my more or less reusable code while I'm on the road (or, at least, using different machines).

Stratified Sample Design For Multilevel Models

The sampling design simulation software can be downloaded from here .

Plotting

• Plot medians with line segments representing 95% confidence intervals to compare the sampling distribution of a statistic across 4 variables and within each variable across 3 conditions. Confidence Segment Plot
  
• Plot periodograms with confidence intervals calculated for 1) the null hypothesis of white noise, and 2) the smoothed periodogram values Spectrum Plots (for R) .

Analysis

• Functions for pooled regression analysis with appropriate weights with finite sample correction. Also a Wald test function for linear combinations of coefficients or other joint tests. Pooled Regression and Wald Test [Note : John Fox recently released a library for R that contains a more polished wald test --- see car ]
• Generate datasets to represent a particular multilevel model, estimate a linear mixed-effects model (aka hierarchical linear model, random coefficients model,etc..), save the results in a comma-delimited data file. Do this 1000 times to produce a sampling distribution for a variety of statistics. See the Sample Design website for more description of this software. LME Simulation
• Natural Cubic B-Splines within multilevel models (under construction).

Data Manipulation